A comprehensive, production-grade image processing toolkit powered by state-of-the-art AI technology. Perform professional background removal, watermark removal, and intelligent image splitting with complete privacy and control.
- Ultra Backend - Maximum quality with BRIA RMBG-2.0 (non-commercial)
- Advanced Processing - Trimap refinement, multi-space defringing
- Color Optimization - Special handling for green/white/black backgrounds
- Controllable Strength - Fine-tune removal aggressiveness (0.1-1.0)
- Gemini Watermark Remover - Remove Google Gemini AI watermarks
- Reverse Alpha Blending - Mathematical precision removal
- Auto Detection - Automatic watermark size detection (48px/96px)
- Adjustable Strength - Partial or complete removal
- Smart Grid Layout - Automatic optimal grid calculation
- Custom Arrangements - Flexible rows × columns configuration
- Quality Preservation - No compression, lossless splitting
- Batch Processing - Split multiple images at once
Requirements: Python 3.13+ and uv package manager
git clone https://github.com/yourusername/Image-Tools.git
cd Image-Tools
uv syncuv run main.pyFollow the interactive prompts:
- Select Folder - Choose input directory
- Select Operation - Background removal, watermark removal, or image splitting
- Configure Settings - Adjust parameters for your use case
- Process - Let the tool handle the rest
Output files are saved in <input-folder>/output/
The ultimate solution for users seeking the absolute best quality:
Features:
- BRIA RMBG-2.0 - Professional-grade model with superior training data
- Trimap Refinement - Intelligent boundary processing (preserves details)
- Multi-space Defringing - RGB + LAB + HSV color analysis
- Guided Filter - Edge-aware smoothing (better than Gaussian blur)
- Color Filter - Optimized for pure-color backgrounds
| Scenario | Strength | Color Filter |
|---|---|---|
| General photos | 0.6-0.8 | Off |
| Complex edges (hair/fur) | 0.7-0.9 | Off |
| Green screen | 0.8-0.9 | Green |
| White background (product) | 0.8-0.9 | White |
| Black background (studio) | 0.8-0.9 | Black |
✅ Enable for:
- Green screen photography/video
- Product photography (pure white/black background)
- ID photos (solid color background)
❌ Disable for:
- Natural scenes (complex backgrounds)
- Gradient backgrounds
- Mixed backgrounds
Remove Google Gemini AI watermarks using reverse alpha blending algorithm.
Features:
- Automatic Detection - Auto-detect watermark size based on image dimensions
- Manual Override - Force 48px or 96px mode if needed
- Strength Control - Adjust removal intensity (0.1-1.0)
- Precision Algorithm - Mathematical reverse alpha blending
Detection Rules:
- Images > 1024×1024 pixels → 96×96 watermark
- Images ≤ 1024×1024 pixels → 48×48 watermark
Split large images into smaller tiles with intelligent layout.
Features:
- Smart Grid - Auto-calculate optimal rows × columns
- Custom Layout - Specify exact grid configuration
- Quality Preservation - Lossless PNG output
- Batch Processing - Process multiple images
Use Cases:
- Social media carousels (Instagram, Twitter)
- Print layouts and posters
- Game tile maps
- Large artwork segmentation
Input Image
↓
Stage 1: BRIA RMBG-2.0 Segmentation
├─ Professional-grade alpha matte generation
├─ High-resolution processing (1024×1024)
└─ Strength-based threshold adjustment
↓
Stage 2: Trimap Refinement (Optional)
├─ Identify uncertain boundary regions
├─ Guided filter for edge-aware smoothing
└─ Preserve fine details (hair, fur)
↓
Stage 3: Multi-space Defringing
├─ RGB color balance analysis
├─ LAB color space correction
└─ Alpha-based edge blending
↓
Stage 4: Color Filter (Optional)
├─ HSV/LAB color space detection
├─ Morphological mask refinement
├─ Edge despill (for green screens)
└─ Alpha channel merging
↓
Output PNG (RGBA)
- Single Image: 2-5 seconds (depends on resolution and hardware)
- Memory Usage: ~3-4GB GPU (BRIA RMBG-2.0), ~2GB CPU
- Batch Processing: Efficient session reuse
- Quality: Professional-grade, comparable to commercial services
| Configuration | Specs |
|---|---|
| Minimum | 8GB RAM + CPU (slower) |
| Recommended | 16GB RAM + 4GB+ VRAM GPU (CUDA) |
| Optimal | 32GB RAM + 8GB+ VRAM GPU |
| Feature | Commercial Tools | Image-Tools |
|---|---|---|
| Privacy | ❌ Upload required | ✅ 100% local |
| Customization | ❌ Limited control | ✅ Full control |
| Cost | 💰 $0.20+/image | ✅ Free (hardware only) |
| Quality | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| Features | Background removal only | ✅ Multi-tool suite |
| Commercial Use | ✅ Allowed (paid) |
All dependencies are open-source with permissive licenses:
- transformers (Apache 2.0) — HuggingFace transformers for BRIA RMBG-2.0
- torch (BSD-3) — PyTorch deep learning framework
- opencv-python (Apache 2.0) — Computer vision operations
- pillow (HPND) — Image processing utilities
- InquirerPy (MIT) — Interactive CLI interface
- numpy (BSD-3) — Numerical computing
This project builds upon excellent open-source work:
- BRIA RMBG-2.0 by BRIA AI - Professional-grade model
- BiRefNet by ZhengPeng7 - SOTA segmentation research
- Gemini Watermark Remover by journey-ad - Watermark removal algorithm
- Cloudflare - For evaluating background removal models
MIT License - See LICENSE for details
This means:
- ✅ Personal use is allowed and free
- ✅ Research and educational use is allowed
- ❌ Commercial use is NOT allowed without separate licensing
- For commercial use, please contact BRIA AI for licensing
If you find this project useful, please consider giving it a star ⭐
Contributions are welcome! Please feel free to submit a Pull Request.
For questions, issues, or feature requests, please open an issue on GitHub.
Made with ❤️ for the open-source community